Comprehensive Study Conducted on the Potential of Seismic Events in Coal Mining Procedure

Accurate prediction of seismic events is critical to enhancing safety and mitigating risks in coal mining operations. This study uses the long-term correlations present in historical seismic data to investigate how well machine learning (ML) techniques forecast seismic activity. Specifically, an Extra Trees Classifier (ETC) is employed and enhanced through integration with two metaheuristic optimization algorithms: Red-Tailed Hawk (RTH) and Bonobo Optimization (BO). These hybrid models, ETRT and ETBO, are developed to improve prediction precision under both hazardous and non-hazardous conditions. Results demonstrate that while the base ETC model achieves high accuracy in hazardous scenarios (precision: 0.990 ), the hybrid models outperform it with a perfect precision of 1.000 . Under non-hazardous conditions, the ETBO scheme shows improved performance over ETC, achieving a precision of 0.780 compared to 0.540 , though the ETRT model remains the most effective overall (precision: 0.870 ). The findings highlight the potential of hybrid ML-optimization frameworks in seismic risk forecasting and contribute to advancing predictive tools for safer mining operations.